Key Comparison. CCQM K Electrolytic Conductivity at 0.5 S m -1 and 5 ms m -1. Final Report

Transcription

1 Key Comparison CCQM K Electrolytic Conductivity at 0.5 S m -1 and 5 ms m -1 Final Report Steffen Seitz, Beatrice Sander Alan Snedden, Lisa DeLeeBeeck, Galia Ticona Canaza, Toshiaki Asakai, Igor Maksimov, Xiaoping Song, Hai WANG, Wladyslaw Kozlowski, Joanna Dumanska, Beáta Jakusovszky, Zsófia Nagyné Szilágyi, Vladimir Gavrilkin, Olexiy Stennik, Yuri Ovchinnikov, Simone Fajardo Ferraz, Fabiano Barbieri Gonzaga, Kleiton da Cruz Cunha,Zuzana Hanková, Michal Máriássy, Martina Vicarova, Alena Vospelova, Jose Luis Ortiz-Aparicio, Judith Velina Lara-Manzano, Jorge Uribe-Godínez, Daniela Stoica, Paola Fisicaro, V.I.Suvorov, Leonid A Konopelko, Aleksey M.Smirnov, Ronald Cristancho Amaya, Henry Torres Quezada Summary Key Comparison CCQM-K was a follow-up comparison for K36 and provided updated support for the corresponding calibration and measurement capability (CMC) entries in the BIPM CMC database. It aimed to demonstrate the capabilities of the participating NMIs to measure electrolytic conductivity of aqueous electrolyte solutions in the conductivity range 0.15 S m -1 to 1.5 S m -1 and in the conductivity range 1.5 ms m -1 to 15 ms m -1. To this end electrolytic conductivity of a potassium chloride solution (nominal conductivity 0.5 S m -1 ) and of a HCl solution (nominal conductivity 5 ms m -1 ) had to be measured. 17 NMIs participated in the comparison. The key comparison reference value (KCRV) of the KCl solution was ( / ) S m -1 and the KCRV of the HCl solution was ( /-0.012) ms m -1. Both values were estimated from the medians of the results considered eligible for KCRV calculation. They were given with their expanded uncertainties (95% coverage). The majority of the 0.5 S m -1 results were consistent with the KCRV. Two institutes showed a small inconsistency, one outlier was observed. The conductivity of the HCl solution showed a small, but steady linear drift of ms m -1 per day during the measurement period that was corrected for KCRV calculation. Some institutes reported unstable measurement conditions for this solution. The results of seven participants have been inconsistent with the KCRV

2 CCQM-K Content Coordinating laboratory and contact persons... 3 Metrology Area... 4 Branch... 4 Subject... 4 Time schedule... 4 Description of samples... 5 Preparation, shipment, handling... 5 Homogeneity... 7 Stability... 8 Correspondence with institutes Results of 0.5 S m -1 KCl solution Reported results KCRV calculation Degrees of Equivalence Results of 5 ms m -1 HCl solution Reported results KCRV calculation Degrees of Equivalence Effect of Carbon Dioxide How Far Does The Light Shine statement Traceability statement References Appendix A: Equations used to calculate KCRV estimates / 24

3 CCQM-K Coordinating laboratory and contact persons Physikalisch-Technische Bundesanstalt WG Fundamentals in Electrochemistry and electrochemical Energy Storage Bundesallee 100 GERMANY Braunschweig Steffen Seitz steffen.seitz@ptb.de Phone +49 (0) Fax +49 (0) Beatrice Sander beatrice.sander@ptb.de phone +49 (0) fax +49 (0) / 24

4 CCQM-K Metrology Area Amount of Substance Branch Electrochemistry Subject Measurement of the electrolytic conductivity of two unknown samples with nominal values 0.5 S m -1 (aqueous KCl solution) and 5 ms m -1 (aqueous HCl solution. Time schedule Invitation June 2016 Registration Deadline 31 July 2016 Sample preparation August/September 2016 Sample shipment Beginning of November 2016 Reporting Deadline 27 January 2017, extended Draft A April 2017 Draft B report April 2018 List of participants and contact persons No Akkr Institute Country 1 DFM 2 INACL 3 NMIJ 4 NIM 5 GUM 6 BFKH 7 UkrCSM 8 VNIIFTRI 9 LATU Dansk Fundamental Metrologi DEN Instituto Nacional de Calidad PER National Metrology Institute of Japan National Institute of Metrology (NIM) Central Office of Measures (GUM) Government Office of the Capital City Budapest State Enterprise All- Ukrainian State Research and production Center of Standardization Metrology Federal Agency on technical regulation and metrology of Russia Laboratorio Tecnológico del Uruguay JPN CHN POL HUN UKR RUS URY contact person Alan Snedden asn@dfm.dk Galia Ticona Canaza gticona@inacal.gob.pe Toshiaki Asakai t-asakai@aist.go.jp Xiaoping Song songxp@nim.ac.cn Wladyslaw Kozlowski w.kozlowski@gum.gov.pl Beáta Jakusovszky jakusovszky@mkeh.hu Vladimir Gavrilkin vgavrilkin@ukrcsm.kiev.ua Yuri Ovchinnikov jao@vniiftri.ru Simone Fajardo Ferraz sfajardo@latu.org.uy 4 / 24

5 CCQM-K No Akkr Institute Country 10 INMETRO 11 SMU Instituto Nacional de Metrologia Qualidade e Tecnologia Slovak Institute of Metrology BRA SVK 12 CMI Czech Metrology Institute CZE 13 CENAM 14 PTB 15 LNE 16 VNIIM 17 INM Centro Nacional de Metrología Physikalisch-Technische Bundesanstalt Laboratoire National de Métrologie et d Essais D.I.Mendeleyev Institute for Metrology National Metrology Institute of Colombia MEX GER FRA RUS COL contact person Fabiano Barbieri Gonzaga fbgonzaga@inmetro.gov.br Zuzana Hanková hankova@smu.gov.sk Martina Vicarova mvicarova@cmi.cz Jose Luis Ortiz-Aparicio jortiz@cenam.mx Beatrice Sander beatrice.sander@ptb.de Daniela Stoica daniela.stoica@lne.fr V.I.Suvorov V.I.Suvorov@vniim.ru Ronald Cristancho Amaya rcristancho@inm.gov.co Description of samples Preparation, shipment, handling The solutions used for the comparison have been produced by the coordinating laboratory. 20 L KCl solution with a nominal conductivity value of 0.5 S m -1 have been prepared in a 25 L PE vessel using potassium chloride (Suprapur from Merck Millipore) and pure water. The water has been taken from a Millipore Elix purification system. The solution has been aerated for several hours after preparation to achieve equilibrium condition with ambient air. Afterwards the KCl solution has been homogenized, filled in 200 ml bore silicate glass bottles and closed with rubber stoppers. The stoppers have been fixed with aluminium crimps. Afterwards they have been weighed. 20 L HCl solution with a nominal conductivity of 5 ms m -1 has been prepared by dilution of hydrochloric acid (Titrisol 0.01 M from Merck Millipore) with pure water in a 25 L PE vessel. The solution has been aerated after preparation for a few days to achieve equilibrium condition with ambient air. In the meanwhile 500 ml HDPE bottles have been filled and stored with a similar HCl solution for a few days. Afterwards they have been spilled once with pure water and dried in an oven at 60 C. Then the HCl solution for the comparison has been homogenized and filled into the bottles. The bottles have been closed with lids immediately after filling. The lids have been sealed with tape and then the bottles have been weighed. Afterwards they have been sealed in aluminium laminated bags to prevent water evaporation. The samples have been dispatched to all participants at the same time. They were shipped in a cardboard box by courier. After receipt the bottles have been visually inspected for damages. Bottle masses have been measured and corrected for air buoyancy. The results have been reported to the coordinating institute, where they have been compared with the initially weighed masses in order to verify bottle integrity. DFM and VNIIFTRI bottles have shown significantly deviating masses for the HCl solutions (also see section communication with institutes). One bottle of KCl solution sent to VNIIFTRI showed a deviation of about 0.1 g. Mass 5 / 24

6 mass difference /g CENAM CMI CMI DFM DFM GUM GUM INACAL INM Inmetro LATU LATU LNE LNE BFKH BFKH BFKH NIM NMIJ NMIJ SMU UkrCSM VNIIFTRI VNIIM VNIIM CCQM-K differences of all other bottles were smaller than 0.05 g (see figures 1 and 2). Apart from the conspicuous VNIIFTRI values all other masses of the HCl solution bottles have decreased. A plot of the deviation against the date of weighing (not shown) shows a tendency to decreasing masses and to increasing spread. Apparently, there has been a small residual loss of water. However, since the drop was % on average its effect on conductivity could be neglected. After weighing the bottles had to be stored below 25 C until the measurements started. The bottles containing HCl solution had to be put back into the aluminium bags for storage until the measurement, whereas the bags had to be closed with tape, but not to be re-sealed Figure 1 Measured mass differences of 0.5 S/m KCl solution with respect to the initial weighing at the coordinating institute. One value from VNIIFTRI has been cut off. 6 / 24

7 mass differences /g CENAM CMI DFM DFM GUM GUM INACAL INM Inmetro LATU LATU LNE LNE LNE BFKH BFKH NIM NMIJ NMIJ SMU UkrCSM UksCSM VNIIFTRI VNIIFTRI VNIIM CCQM-K Figure 2 Measured mass differences of 5 ms/m HCl solution with respect to the initial weighing at the coordinating institute. The values from VNIIFTRI have been cut off. Homogeneity 0.5 S m -1 KCl solution 5 bottles have been chosen for homogeneity testing with bottle numbers distributed over the whole batch. A Guildline Autosal B salinometer has been used to measure conductances of the samples. The relative standard deviation of the conductances was %. This is significantly smaller than the measurement uncertainties of conductivity measurements, so that the samples can be assumed sufficiently homogeneous for the comparison. The results of the individual measurements are shown in table 1. Table 1 Results of homogeneity test of the KCl solution samples bottle # bath temperature conductance C arbit. units Mean conductance (arbit. units): Standard deviation < Relative standard deviation % 7 / 24

8 CCQM-K ms m -1 HCl solution 5 bottles have been chosen for homogeneity testing with bottle numbers distributed over the whole batch. A two electrode Jones-type cell has been used to measure conductivity of the samples. The relative standard deviation was %. This is significantly smaller than the expected measurement uncertainties of the comparison conductivity measurements, so that the samples can be assumed sufficiently homogeneous for the comparison measurement. The results of the individual measurements are shown in table 2. Note that conductivities have been linearly corrected to C. Table 2 Results of the homogeneity test of the HCl solution samples bottle # bath temperature C C ms/m Mean conductivity: ms/m Standard deviation: ms/m Relative standard deviation % Stability 0.5 S m -1 KCl solution 5 bottles have been arbitrarily chosen for stability testing. A Guildline Autosal B salinometer has been used to measure conductances of the samples. Measurements have been conducted in approximately 4 week intervals over the whole measurement period. No significant drift has been observed. The relative standard deviation is %, which is small enough for the requirements of this comparison, albeit a factor 3 larger than the measured homogeneity. Therefore another homogeneity test has been performed at the end of the measurement period. The relative standard deviation of that measurement was %, which was even an order of magnitude smaller than the first homogeneity test. The relative deviation of the means of both homogeneity tests is %. Thus the spread of the stability results is more likely due to (small) instabilities of the measurement device during the measurement period than due to sample instabilities. Consequently the KCl solution samples can be assumed sufficiently stable for the comparison measurement. The results of the individual measurements are shown in table 3 and figure 3. 8 / 24

9 conductance /arbit. units CCQM-K Table 3 Results of the stability test of the KCl solution samples bottle # date measured bath temperature C C arbit. units from homgeinty tests , , 35, 57, 78, Mean conductance (arbit. units): Standard deviation: Relative standard deviation % days Figure 3 Stability tests of KCl solution samples. The error bars indicate the double standard deviation of the (first) homogeneity test. Note that the smallest, expanded standard uncertainties of the comparison are about a factor 10 larger. 9 / 24

10 CCQM-K ms m -1 HCl solution 4 bottles have been arbitrarily chosen for stability testing. The samples have been stored in the laboratory at room temperature and kept in the aluminium bags until the measurement. A two electrode Jones-type cell has been used to measure conductivity of the samples. Measurements have been conducted in approximately 4 week intervals over the whole measurement period. A linear increase of the measured conductivity values of ms m -1 per day has been observed. The results of the individual measurements are shown in table 4 and figure 4. Note that conductivities have been linearly corrected to C. After drift had been realised stability of the Jones-type cell has been verified with a stable 100 ms m -1 KCl solution to check if the measured drift results from a change of solution conductivity, but not from a change of cell properties. The samples have been prepared and characterised for homogeneity in the same way as the 0.5 S m -1 KCl solution of this comparison. Starting , KCl solution samples have been measured once a month during the remaining measurement period. Since the spread of the resistance values over time (<0.004%) was compatible with the results of the homogeneity measurement, it is concluded that the cell and the 100 ms m -1 KCl test solution were stable during the measurement period. At the end of the measurement period the homogeneity test has been repeated with another 4 bottles of HCl solution. The relative standard deviation of those measurements has been 0.01 %, which is somewhat larger than the spread of the first homogeneity test, however, it is still acceptable. Thus it can be reasonably assumed that all HCl solution samples underwent a similar drift within acceptable limits. All reported measurement results have been corrected for the observed drift to a date in the middle of the measurement period ( ) in order to calculate the KCRV and the DoEs. It must be noted that the drift correction had only a small effect on the results. Table 4 Results of the stability test of the HCl solution samples bottle # date measured bath temperature C from homogeneity tests C arbit. units (mean) , (mean) ,8,31,36, (mean) / 24

11 conductivity /(ms/m) CCQM-K y = E-05x E days Figure 4 Stability tests of the HCl solution samples. The error bars indicate the double standard deviation of the (first) homogeneity test. Note that measurement uncertainties of most institutes are significantly larger. Correspondence with institutes DFM has been informed that the reported bottle masses of the HCl solution deviated significantly from the masses measured at the coordinating laboratory. It turned out that DFM had reported masses including the aluminium bag and has sent results without the bags later on, which then showed a deviation within the accepted limits. VNIIFTRI has reported a significantly delayed delivery of the samples ( ) due to problems with customs clearance. As a consequence, VNIIFTRI could not meet the reporting deadline. When the report finally arrived end of February a few days after the extended deadline, there was no time for further actions. However, the deviation regarding the HCl solution bottle masses matched the mass of the aluminium bag, so VNIIFTRI most probably didn t remove the bag either. Moreover, the reported conductivity results of the two KCl solution samples measured by VNIIFTRI didn t correspond to the observed mass differences. In contrast, the conductivities of both samples were equivalent within measurement uncertainty. Consequently, integrity of all samples sent to VNIIFTRI can be assumed. BFKH and CMI have reported problems to meet the deadline due to illness of staff. INMETRO has reported technical problems. The coordinating institute has agreed to postpone the deadline until Eventually all reports have arrived, even though with an acceptably small delay. 11 / 24

12 CCQM-K Conducting a preliminarily evaluation the coordinating laboratory identified results showing unusual deviation of some measurement results from the median or unusual measurement uncertainties. Moreover, some of the reports did not contain requested information. The institutes concerned were asked to provide the lacking information and to check their results for calculation or typing errors respectively. No information was given about the magnitude or the sign of the deviation. Subsequently the following institutes have sent revised values: CENAM, GUM, INM, INMETRO, VNIIFTRI, UkrCSM, LATU and CMI. LNE and NMIJ have confirmed the original results. UkrCSM, VNIIFTRI, VNIIM and CMI have sent additional information. Results of 0.5 S m -1 KCl solution Reported results Table 5 lists the reported results. The last column lists the stated source of traceability. Figure 5 shows the results graphically. Table 5 Reported conductivity results of the 0.5 S m -1 KCl solution Laboratory i quantity value i standard uncertainty u( i) coverage factor k i expanded (95%) uncertainty U( i) cell type source of traceability akkr. S/m S/m S/m VNIIM secondary VNIIM RM / recipe OIML NMIJ primary/ Jones SI GUM primary/ Piston SI INMETRO primary/ Piston SI INACAL secondary SMU CRM LATU secondary DFM CRM VNIIFTRI secondary VNIIFTRI RM/SI BFKH secondary PTB / 24 primary/ Piston BFKM RM / recipe/ OIML SMU secondary SMU RM/recipe Bradshaw values DFM secondary DFM CRM / SI NIM secondary NIM RM / recipe IUPAC CMI secondary CMI CRM / SI LNE primary/ Jones SI UkrCSM secondary Ukraine RM / SI CENAM secondary SMU CRM INM secondary SMU CRM SI

13 VNIIM NMIJ GUM INMETRO INACAL LATU VNIIFTRI BFKH PTB SMU DFM NIM CMI LNE UkrCSM CENAM INM conductivity of KCl solution /(S/m) CCQM-K Figure 5 Reported results of the 0.5 S/m KCl solution. The error bars indicate standard uncertainties. KCRV calculation Only independent results have been used for calculation of the key comparison reference value (KCRV) [1]. Hence, values of institutes that have used standards from other participating institutes for cell calibration have been excluded from KCRV calculation. Three candidate KCRV estimates and the corresponding standard uncertainties have been calculated (see appendix A for the formula used): the arithmetic mean, the (external) weighted mean and the median. The candidate KCRVs and their standard uncertainties are shown in table 6. A Chi-square consistency check of the input data for the weighted mean calculation has been performed according to [2]. Since the F13 value is larger than the X12,0.95 value the results must be considered inconsistent with the weighted mean. Because of the deviating VNIIM value, the arithmetic mean does not reflect the bulk of the results and it has an unacceptably large uncertainty. EAWG has decided in the CCQM spring meeting 17. April 2017 in Paris to use the median as KCRV. Even though the result of VNIIM is probably an outlier (for unknown reasons) it has not been excluded, since its effect on the median is negligible. Hence, the value of the chosen KCRV and its standard uncertainty is KCRV(0.5 S m -1 KCl) = ( ± ) S m / 24

14 conductivity /(S/m) CCQM-K Table 6 Candidate KCRVs of the 0.5 S/m KCl solution and their standard uncertainties. weighted mean WM u WM F13( WM) X12,0.95 S/m S/m ,03 results are inconsistent with the weighted mean u M median M S/m S/m mean µ Std.-Dev S/m S/m weighted mean median mean Figure 6 Candidate KCRVs for the 0.5 S/m KCl solution. The error bars indicate their expanded uncertainties. Degrees of Equivalence Table 7 shows the degrees of equivalence (DoE) with the KCRV. Their expanded uncertainties, shown in column 4, have been calculated from the combined expanded uncertainties of the reported values and the KCRV (95% coverage). The last column indicates the minimal standard uncertainties um(i) for laboratory i that are consistent with the KCRV. They can be stated for CMC entries. In case of consistency of a result with 14 / 24

15 CCQM-K the KCRV, that is DoEi U(DoEi), um(i) is the reported standard uncertainty. If it is not consistent with the KCRV um(i) is calculated with [3] 2 2 DoE 2 2 u ( ) i m i u ( KCRV ) ki k. (1) i ki is the reported coverage factor. The minus sign applies for results that contributed to the KCRV calculation and the plus sign applies to those that didn t contribute. Figure 7 shows the DoEs and their uncertainties graphically. Table 8 Degrees of Equivalence for the 0.5 S/m KCl solution lab i DoE i lab contributed to KCRV U(DoE i) consistency DoE i U(DoE i) E n um( i) minimal (CMC) uncertainty consistent with KCRV akkr.. S/m yes/no S/m yes/no DoE i/u(doe i) VNIIM yes no NMIJ yes yes GUM yes no INMETRO yes yes INACAL no yes LATU no yes VNIIFTRI yes yes BFKH yes yes PTB yes yes SMU yes yes DFM yes yes NIM yes yes CMI yes yes LNE yes yes UkrCSM yes no CENAM no yes INM no yes / 24

16 VNIIM NMIJ GUM INMETRO INACAL LATU VNIIFTRI BFKH PTB SMU DFM NIM CMI LNE UkrCSM CENAM INM Degree of Equivalence / (S/m) CCQM-K Figure 7 Degrees of equivalence and the corresponding expanded uncertainties. Results of 5 ms m -1 HCl solution Reported results Table 9 lists the reported results for the 5 ms m -1 HCl solution. The third column shows the results corrected for drift as described above. Figure 8 shows the reported results and the drift corrected results graphically. Obviously, the conductivity drift during storage was small and the effect on the KCRV and the DoEs is insignificant compared to the uncertainty of the KCRV (0.03% vs 0.12%). Some institutes have measured more than one bottle at different dates. In this case a mean date has been used for drift correction (7 th column). The 8 th column lists times that have passed between filling of the cells and the actual measurements. The 9 th column lists if the corresponding institute has corrected for a short-term drift during the measurement. 16 / 24

17 CCQM-K Table 9 Reported conductivity results of the 5 ms m -1 HCl solution lab i quantity value i quantity value i (drift corrected) standard uncertainty u( i) coverage factor k i expanded (95%) uncertainty U( i) date measured (mean) time between filling and measurement abbr. ms/m ms/m ms/m ms/m min drift during measurement corrected cell (*) source of traceability NMIJ n/m primary SI INM no secondary SMU CRM GUM yes primary SI CMI n/m secondary CMI RM/SI SMU no secondary SMU CRM / recipe Bradshaw values INACAL n/m n/m secondary SMU CRM BFKH n/m secondary BKEH RM/recipe OIML NIM no secondary NIM RM/recipe IUPAC LNE n/m primary SI DFM yes secondary DFM RM/SI PTB yes secondary PTB RM / SI UkrCSM yes secondary UkrCSM RM/SI LATU n/m secondary DFM CRM INMETRO yes primary SI VNIIM n/m secondary VNIIM RM/recipe OIML CENAM n/m secondary SMU CRM VNIIFTRI n/m secondary VNIIFTRI RM/SI n/m = not mentioned (*) for the kind of primary cell type see table 5 17 / 24

18 NMIJ INM GUM CMI SMU INACAL BFKH NIM LNE DFM PTB UkrCSM LATU INMETRO VNIIM CENAM VNIIFTRI conductivity of HCl solution (ms/m) CCQM-K reported 5.06 drift corrected Figure 8 Results of the 5 ms/m HCl solution. The solid diamonds indicate the results corrected for (storage) drift, while the void squares show the originally reported values. The error bars indicate standard uncertainties. KCRV calculation Three candidate KCRV estimates and the corresponding standard uncertainties have been calculated (see appendix A for the formula used): the arithmetic mean, the (external) weighted mean and the median. The results are shown in table 6. A Chi-square consistency check of the input data for the weighted mean calculation has been performed [2]. Since the F13 value is larger than the X12,0.95 value the results must be considered inconsistent with the weighted mean. Basically, all candidates provide compatible KCRV estimates, whereas the median is more robust with respect to outliers. Therefore, CCQM-EAWG has decided in the meeting held 17. April 2017 in Paris to use the median as KCRV. Hence, the value of the chosen KCRV and its standard uncertainty is KCRV(5 ms m -1 HCl) = ( ± ) ms m / 24

19 conductivity /(ms/m) CCQM-K Table 10 Estimates for the KCRV of the 5 ms/m HCl solution weighted mean WM u WM F13( WM) X12,0.95 ms/m ms/m results are inconsistent with the weighted mean u M median M ms/m ms/m mean µ Std.-Dev ms/m ms/m weighted mean median mean Figure 9 Various candidate KCRV estimates for the 5 ms/m HCl solution. The error bars indicate their expanded uncertainties (k=2). Degrees of Equivalence Table 11 shows the degrees of equivalence based on the median. Their expanded uncertainties, shown in column 4, have been calculated from the combined expanded uncertainties of the reported values and the KCRV (95% coverage). The last column indicates the minimal uncertainties um(i) that are consistent with the KCRV. They can be stated for CMC entries. In case of consistency with the KCRV um(i) is the reported standard uncertainty. In case it is not consistent with the KCRV it is calculated with eq (1). Figure 10 shows the DoEs and their uncertainties graphically. 19 / 24

20 CCQM-K Table 11 Degrees of equivalence for the 5 ms/m HCl solution lab lab i DoE i contributed to KCRV U(DoE i) consistency DoE i U(DoE i) E n um( i) akkr. S/m yes/no S/m yes/no DoE i/u(doe i) minimal (CMC) uncertainty consistent with KCRV NMIJ yes no INM no no GUM yes no CMI yes yes SMU yes yes INACAL no yes BFKH yes yes NIM yes yes LNE yes yes DFM yes yes PTB yes yes UkrCSM yes no LATU no yes INMETRO yes no VNIIM yes no CENAM no yes VNIIFTRI yes no / 24

21 NMIJ INM GUM CMI SMU INACAL BFKH NIM LNE DFM PTB UkrCSM LATU INMETRO VNIIM CENAM VNIIFTRI Degrees of Equivalence /(ms/m) CCQM-K Figure 10 Degrees of equivalence and the corresponding expanded uncertainties of the 5 ms/m HCl solution. Effect of Carbon Dioxide HCl has been chosen for this comparison to supress the effect of CO2 on the conductivity measurement result. Aqueous solutions not preventing the dissociation of dissolved CO2, such as commonly used aqueous KCl and NaCl conductivity standards, can be subject to significant conductivity variation in the conductivity range below 15 ms m -1 if exposed to ambient air. Common measurement practice does not avoid such air contact, since this would require immense efforts. The contribution of dissolved and subsequently dissociated CO2 to the conductivity of such solutions is around 0.1 ms m -1 under equilibrium conditions [4]. Hence, the uncertainty budget must account for this kind of solution instability. In fact, it is only the change of the amount of hydrogen carbonate and carbonate ions during the measurement process that contributes to the uncertainty. This uncertainty contribution is difficult to be estimated in general since the change depends on the partial CO2 pressure present at the site of measurement, the usually unknown carbonate ion concentrations in the solution and the time the solution is exposed to air during filling and measurement. Consequently, it might be negligible in one measurement and substantial in another. However, under typical operation procedures it is reasonable to assume that the conductivity change due to CO2 is not larger than ±0.02 ms m -1 with respect to the initial value before opening the bottle. Thus, a standard uncertainty contribution of ms m -1 should be generally considered in the combined standard uncertainty of conductivity measurements of pure aqueous electrolyte solutions. 21 / 24

22 E n values of HCl results E n values of KCl results CCQM-K How Far Does The Light Shine statement The results of this KC are considered representative for conductivity measurement capabilities of stable aqueous electrolyte solutions in the conductivity range 0.15 S m -1 to 1.5 S m -1 and 1.5 ms m -1 to 15 ms m -1, respectively. For the reasons described in the previous section an uncertainty contribution of ms m -1 must be considered in the minimal uncertainty umco2(i) permissible for CMCs of aqueous solutions, having conductivities 15 ms m -1, in order to account for the effect of CO2 on conductivity: u mco2 (κ i ) = u m2 (κ i ) + (0.012 ms um(i) corresponds to the value of laboratory i given in table 11. m )2 (2) This statement particularly holds for aqueous, low conductivity KCl and NaCl conductivity standard solutions. Traceability statement During the spring meeting in 2016 EAWG has decided to accept IUPAC Technical Report 2001 [4] as metrological reference for the traceability of conductivity measurement results for practical reasons. Therein conductivity values and their uncertainties are assigned to KCl solutions, prepared according to a defined recipe. An evaluation of previous key comparisons and pilot studies has shown no evidence for a significant deviation in the DoEs of such results compared to those traceable to the SI. It was also decided that this statement has to be verified based on the results of every future conductivity key comparison in order to retain the link of such measurements to the SI. Figure 9 shows the normalised DoEs of all results contributing to the KCRV. No significant deviations of recipe based values are observed. Hence, the statement is still considered valid arbitrary number arbitrary number Figure 11 En values (DoEs normalized with the corresponding U(DoE)). Solid diamonds indicate results traceable to the SI, void diamonds those traceable to [4]. En 1 indicates consistency with the KCRV. 22 / 24

23 CCQM-K References [1] M. G. Cox, "The evaluation of key comparison data," Metrologia, vol. 39, pp (2002). [2] M. G. Cox, "The evaluation of key comparison data: determining the largest consistent subset," Metrologia, vol. 44, pp (2007). [3] M.G. Cox, P. Harris, M. Milton, Method for determining acceptable CMCs to ensure consistency with KC results, 2009, unpublished [4] K. W. Pratt, W. F. Koch, Y. C. Wu, and P. A. Berezansky, "Molality-based primary standards of electrolytic conductivity," Pure Appl. Chem., vol. 73, pp (2001). 23 / 24

24 CCQM-K Appendix A: Equations used to calculate KCRV estimates n i i ui number of labs contributing to the estimate = 1.. n, index for lab contributing to the estimate conductivity value of lab i standard uncertainty of i Arithmetic mean κ = κ i n Standard deviation of arithmetic mean i n n 1 2 Weighted mean WM w i i with w i 1 ui 1 ui 2 2 Uncertainty of weighted mean u WM w ( i i ( n 1) WM w i ) Uncertainty of median um Median ( i Median ) n 1 24 / 24